update

README.md

@@ -1,8 +1,8 @@
 ---
-title: Llama 3 Youko 8b Int
-emoji: 🐢
-colorFrom: pink
-colorTo: pink
+title: Llama 3 Youko 8b Instruct
+emoji: 🦊
+colorFrom: yellow
+colorTo: yellow
 sdk: gradio
 sdk_version: 4.29.0
 app_file: app.py

app.py

@@ -0,0 +1,152 @@
+import os
+import gradio as gr
+import spaces
+from transformers import AutoTokenizer, AutoModelForCausalLM, TextIteratorStreamer
+from threading import Thread
+
+
+# Set an environment variable
+HF_TOKEN = os.environ.get("HF_TOKEN", None)
+
+
+model_id = "rinna/llama-3-youko-8b-instruct"
+
+
+DESCRIPTION = """
+<div>
+<p>🦊 <a href="https://huggingface.co/rinna/llama-3-youko-8b-instruct"><b>Llama 3 Youko 8B Instruct</b> (rinna/llama-3-youko-8b-instruct)</a>は、<a href="https://rinna.co.jp">rinna株式会社</a>が<a href=https://huggingface.co/meta-llama/Meta-Llama-3-8B>Meta Llama 3 8B</a>に日本語継続事前学習およびインストラクションチューニングを行った大規模言語モデルです．Llama 3 8Bの優れたパフォーマンスを日本語に引き継いでおり、日本語のチャットにおいて高い性能を示しています。</p>
+<p>🤖 このデモでは、Llama 3 Youko 8B Instructとチャットを行うことが可能です。</p>
+<p>📄 モデルの詳細については、<a href="https://rinna.co.jp/news/2024/07/20240725.html">プレスリリース</a>、および、<a href="https://rinnakk.github.io/research/benchmarks/lm/index.html">ベンチマーク</a>をご覧ください。お問い合わせは<a href="https://rinna.co.jp/inquiry/">こちら</a>までどうぞ。</p>
+</div>
+"""
+
+LICENSE = """
+---
+<div>
+<p>Built with Meta Llama 3</p>
+<p>License: <a href="https://llama.meta.com/llama3/license/">Meta Llama 3 Community License</a><p>
+<p>This space is implemented based on <a href="https://huggingface.co/spaces/ysharma/Chat_with_Meta_llama3_8b">ysharma/Chat_with_Meta_llama3_8b</a>.</p>
+</div>
+"""
+
+PLACEHOLDER = """
+<div style="padding: 30px; text-align: center; display: flex; flex-direction: column; align-items: center;">
+   <img src="https://huggingface.co/rinna/llama-3-youko-8b/resolve/main/rinna.png" style="width: 80%; max-width: 550px; height: auto; opacity: 0.55;  "> 
+   <h1 style="font-size: 28px; margin-bottom: 2px; opacity: 0.55;">Llama 3 Youko</h1>
+</div>
+"""
+
+css = """
+h1 {
+  text-align: center;
+  display: block;
+}
+#duplicate-button {
+  margin: auto;
+  color: white;
+  background: #1565c0;
+  border-radius: 100vh;
+}
+"""
+
+
+# Load the tokenizer and model
+tokenizer = AutoTokenizer.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(model_id, device_map="auto")
+terminators = [
+    tokenizer.eos_token_id,
+    tokenizer.convert_tokens_to_ids("<|eot_id|>")
+]
+
+
+@spaces.GPU(duration=120)
+def chat_llama3_8b(message: str, 
+              history: list, 
+              temperature: float, 
+              max_new_tokens: int
+             ) -> str:
+    """
+    Generate a streaming response using the llama3-8b model.
+    Args:
+        message (str): The input message.
+        history (list): The conversation history used by ChatInterface.
+        temperature (float): The temperature for generating the response.
+        max_new_tokens (int): The maximum number of new tokens to generate.
+    Returns:
+        str: The generated response.
+    """
+    conversation = []
+    conversation.append({"role": "system", "content": "あなたは誠実で優秀なアシスタントです。どうか、簡潔かつ正直に答えてください。"})
+    for user, assistant in history:
+        conversation.extend([{"role": "user", "content": user}, {"role": "assistant", "content": assistant}])
+    conversation.append({"role": "user", "content": message})
+
+    # Need to set add_generation_prompt=True to ensure the model generates the response.
+    input_ids = tokenizer.apply_chat_template(conversation, add_generation_prompt=True, return_tensors="pt").to(model.device)
+
+    print(input_ids)
+    print(tokenizer.decode(input_ids.tolist()[0]))
+    
+    streamer = TextIteratorStreamer(tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True)
+
+    generate_kwargs = dict(
+        input_ids=input_ids,
+        streamer=streamer,
+        max_new_tokens=max_new_tokens,
+        do_sample=True,
+        temperature=temperature,
+        repetition_penalty=1.1,
+        eos_token_id=terminators,
+    )
+    # This will enforce greedy generation (do_sample=False) when the temperature is passed 0, avoiding the crash.             
+    if temperature == 0:
+        generate_kwargs['do_sample'] = False
+        
+    t = Thread(target=model.generate, kwargs=generate_kwargs)
+    t.start()
+
+    outputs = []
+    for text in streamer:
+        outputs.append(text)
+        yield "".join(outputs)
+        
+
+# Gradio block
+chatbot=gr.Chatbot(height=450, placeholder=PLACEHOLDER, label='Gradio ChatInterface')
+
+with gr.Blocks(fill_height=True, css=css) as demo:
+    
+    gr.Markdown(DESCRIPTION)
+    gr.ChatInterface(
+        fn=chat_llama3_8b,
+        chatbot=chatbot,
+        fill_height=True,
+        additional_inputs_accordion=gr.Accordion(label="⚙️ パラメータ", open=False, render=False),
+        additional_inputs=[
+            gr.Slider(minimum=0,
+                      maximum=1, 
+                      step=0.05,
+                      value=0.9, 
+                      label="生成時におけるサンプリングの温度（ランダム性）", 
+                      render=False),
+            gr.Slider(minimum=128, 
+                      maximum=4096,
+                      step=1,
+                      value=512, 
+                      label="生成したい最大のトークン数", 
+                      render=False),
+            ],
+        examples=[
+            ["日本で有名なものと言えば"],
+            ["ネコ: 「お腹が減ったニャ」\nイヌ: 「\nで始まる物語を書いて"],
+            ["C言語で素数を判定するコードを書いて"],
+            ["人工知能とは何ですか"],
+        ],
+        cache_examples=False,
+                     )
+    
+    gr.Markdown(LICENSE)
+    
+
+if __name__ == "__main__":
+    demo.launch()

requirements.txt

@@ -0,0 +1,3 @@
+accelerate
+transformers
+SentencePiece